Context
From discussion #549 (@DRO34): when an alert fires (e.g. high memory usage), there is no follow-up notification when the condition clears. Users have no way to know the alert is resolved without manually checking.
Feature
Add recovery/resolved notifications to the alerting system:
- Track alert state transitions:
firing -> resolved
- When a metric drops back below the threshold (with a configurable hysteresis/cooldown to avoid flapping), send a recovery notification through the same channels as the original alert (email, webhook, etc.)
- Include in the recovery notification: which alert resolved, when it fired, when it resolved, current metric value
Acceptance criteria
Context
From discussion #549 (@DRO34): when an alert fires (e.g. high memory usage), there is no follow-up notification when the condition clears. Users have no way to know the alert is resolved without manually checking.
Feature
Add recovery/resolved notifications to the alerting system:
firing->resolvedAcceptance criteria